AI Is Replacing Researchers — and China’s Private Funds Are Moving Fast

What’s happening

Private asset managers in China are quietly industrializing research with AI agents. It has been reported that financial roles have a theoretical AI-replacement rate as high as 94% (current realized replacement roughly 28%), and managers are treating the gap as an obvious arbitrage: pay one expensive human or run a 24/7 “digital researcher.” OpenClaw (龙虾), an open‑source multi‑agent toolkit, has become emblematic — reportedly used in internal tutorials, roadshows and by buy‑side teams to automate idea generation, factor discovery and report writing. The economics are simple: research salaries in private funds can run into the low millions of RMB annually; if an AI produces usable analysis at a fraction of that cost, what will managers choose?

How the industry is responding

The response is mixed and fast. Some private managers are actively "raising lobsters": online courses teach teams to tame AI agents; Xueqiu Asset Management (雪球资管) and other firms report building internal AI assistants that out‑produce humans on routine tasks, and it has been reported that at least one shop issued new laptops plus a token subsidy to staff specifically for AI experiments. At the same time, large quant shops and veteran technologists are skeptical. Several Shanghai‑based quant practitioners describe OpenClaw as a toy or, more bluntly, a hype cycle — random, non‑systematic and potentially risky for production trading systems. Established quant firms point to deeper, in‑house multi‑agent platforms (for example Apollo AI from Xiyue Investment (喜岳投资)) and say the real edge still requires robust infrastructure and controls.

Why this matters — economics and geopolitics

For private fund bosses the calculus is pragmatic: replace costly, churn‑prone humans with cheaper, always‑on agents that don’t demand carry, travel or vacation. It has been reported that investors such as Howard Marks have framed the issue economically — if an AI can reliably replicate the work of a $200k analyst, the human/AI distinction is secondary to output utility. There is a geopolitical subtext too: Beijing’s push for AI self‑reliance and Western export controls on advanced chips make software and agent architectures an attractive path to productivity gains without new foreign hardware.

The social and market fallout

What’s inevitable is churn. Some jobs will be automated; others will adapt. Fund managers may re‑define research roles as narrower technical tasks, or redeploy people to relationship and strategy roles that agents can’t yet perform. But risks remain: model randomness, security concerns, and widening capability gaps between deep‑pocketed quant firms and smaller players adopting off‑the‑shelf agents. If token costs fall far below “carbon‑based” salaries, will private funds resist a cheap, obedient, trainable researcher? For many in China’s asset management corridor, that is no longer a hypothetical — it’s a corporate strategy.